We've recently encountered two problems with subject numbers. We assign each subject a 5 digit (and digit only - no letters) code, which we enter for any given task.
Problem 1:
Sometimes the output file does not keep the subject number. Through the log file's date and time stamp, we can ascertain which subject's data we are looking at, but the file name has been changed to '1' or '0'. There is no warning that the original number wasn't accepted.

Problem 2:
We recently lost data, because one file over-wrote another. The RAs remember the warning coming up that there was a file with the same name (and they'll never press the 'Yes' button again). The problem is, the code they had typed in wasn't remotely related to the file that was copied over. (The original file was 24035, and the RAs typed in 44036). 24035 was overwritten with 44036's responses, and saved as 24035.

Why might these subject coding errors occur and how can we get them to stop?